Bay Area Restaurant Collective · Panel 2
The Cost of Convenience
Most conversations about AI stay on productivity and innovation. This one is about the infrastructure underneath — data centers, energy, labor, supply chains — and the tradeoffs an operator actually has to weigh. Not fear, not guilt. Just a clearer picture, and a way to check the numbers yourself.
Built for restaurant owners and managers deciding where AI belongs in their business. Every number here links back to its source.
The buildout, in six numbers
Data: Epoch AI, CC-BY- $38B
- capex per GW of IT power
- 2.1 yrs
- typical build time, GW-scale site
- 71%
- of global AI compute owned by 5 hyperscalers
- +15.5
- ECI/yr model capability growth since early 2024
- 7 mo
- doubling time of world AI compute stock
- $370B+
- raised by frontier AI labs

